[gdal-dev] gdalwarp crash

Even Rouault even.rouault at mines-paris.org
Mon Feb 25 01:32:50 PST 2013


Le lundi 25 février 2013 09:35:37, Simon Lyngby Kokkendorff a écrit :
> Hej List,
> 
>   I was using gdal1.10_dev (precompiled from gisinternals.com/sdk) to warp
> a quite large ECW image, but after running for a few days the process
> crashed with an integer overflow error:

When you say "crashed", you mean it exited with the integer overflow error (to 
be opposed as the windows that is displayed by Windows when a process really 
crashes) ?

> 
> C:\Users\simlk\Downloads\release-1600-x64-gdal-mapserver>gdalwarp -s_srs
> "+proj=
> etmerc +ellps=GRS80 +lat_0=0 +lon_0=-39 +k_0= 0.9996 +towgs84=0,0,0,0,0,0,0
> +uni
> ts=m +no_defs" -t_srs "EPSG:4747" -of GTiff -co "BIGTIFF=IF_NEEDED" -ts
> 110000 3
> 00000 -wm 3072 C:\Ortofoto\GR_samlet_ny.ecw T:\TEMP\simlk\out.tiff
> Creating output file that is 110000P x 300000L.
> Processing input file C:\Ortofoto\GR_samlet_ny.ecw.
> 0...10...20...30...40...50...60...70...ERROR 1: Integer overflow :
> nDstXSize=275
> 00, nDstYSize=37500
> 
> I was trying to mimic a utm zone 24 projection in my source srs - and yes I
> did forget the false easting +x_0=500000, but I doubt the problem is
> related to that.

The issue is rather with the -wm 3072. The warping algorithm will currently 
cast a integer overflow error when a memory allocation above 2 GB is attempted. 
So you should try -wm 2047 or less. There's rarely a significant advantage in 
using so big values for -wm.

> Is there some other creation option that I missed for huge
> output?

You could add -co TILED=YES to produce a tiled geotiff. That might help speed 
up things a bit.

> 
> Cheers,
> Simon Kokkendorff, National Geodata Agency of Denmark


More information about the gdal-dev mailing list